A Hybrid Particle Swarm Optimisation with Differential Evolution Approach to Image Segmentation

Abstract

Image segmentation is a key step in image analysis and many image segmentation methods are time-consuming. The Otsu method and Gaussian Mixture Model (GMM) method are popular in image segmentation, but it is computationally difficult to find their globally optimal threshold values. Particle Swarm Optimisation (PSO) is an intelligent search method and has been widely used in many fields. However it is also easily trapped in local optima. In this paper, we propose a hybrid between PSO and Differential Evolution (DE) to solve the optimisation problems associated with the Otsu model and GMM, and apply these methods to natural image segmentation. The hybrid PSO-DE method is compared with an exhaustive method for the Otsu model, and fitted GMMs are compared directly with image histograms. Hybrid PSO-DE is also compared with standard PSO on these models. The experimental results show that the hybrid PSO-DE approach to image segmentation is effective and efficient.